# Results AI Skill Lead Area · Results Block · Decision Map · Oct 23, 2026 (v1.0) --- ## 1. What the Skill Does The Results skill helps teams find out whether a bet paid off. It is the last block in the Lead area of Glare's Decision Map, and the last block of the whole map. Results judge **viability**: do people keep using it, coming back, and paying? They own **Adopt** signals: what people actually do over time, such as retention, repeat use, spend, and adoption. Test results are leading reads that belong to Findings and Bets. Results reads the lagging measures that only exist once the change is live. Results closes the loop: Initiatives and Bets come from Focus, Findings come from Measure, and Outcomes are what Results checks. The skill works in six steps. | Step | What it does | |---|---| | 1. Start from the bet's prediction | The behavior, metric, and window the bet named. No prediction, nothing to check | | 2. Set the baseline | The same metric, same audience, comparable period before the change | | 3. Read Adopt behavior over the window | Did people come back, repeat, stay, or pay? Launch day is not the outcome | | 4. Read up the chain by link type | Use Mapping's link types to decide how strongly to claim the business moved | | 5. Give a verdict | Held, Partly held, Didn't hold, or Too early | | 6. Re-rank the next round | The verdict changes what the business funds next | | Verdict | When to use it | What it does to the priority | |---|---|---| | Held | Adopt behavior and the business KPI moved as predicted, through a link strong enough to claim | Keep funding; scale the bet | | Partly held | Behavior moved, but the chain broke before the business KPI | Keep the priority; fix the broken link | | Didn't hold | The predicted behavior didn't appear | Move the priority down; send the gap back as a hunch | | Too early | The window hasn't closed, or the data is missing | No verdict yet; set the next check date | ### The Window Rule No verdict before the window closes. A lift that fades by day 60 is a different result from one that holds. Strong test results don't change that: until the window closes, the verdict is Too early. ## 2. Business Benefit Without this check, teams report launch-day numbers, call them proof, and move on. Results tells the business which bets to keep funding, which to fix, and which to drop. This helps teams: - find out whether a bet held over time, not just at launch - make claims that match the strength of the evidence - move funding toward priorities that are working - send what didn't hold back to Measure as a new hunch ## 3. Skill Output When used correctly, the skill produces an outcome check. The example below shows how this works for a mobile banking dashboard, 60 days after the bet shipped. | Field | Example Output (Mobile Banking Dashboard) | |---|---| | Bet's Prediction | Sessions ending without an action drop from 38% to under 30% within 60 days of launch | | Baseline | 38%, habitual users, the 60 days before launch | | Window | 60 days after launch (closed) | | Adopt Behavior | Sessions ending without an action: 29%. Weekly return visits held steady | | Up the Chain | Support calls about balance down 22% (Sequential). Churn flat so far (Correlated link; too soon to show) | | Verdict | Partly held: behavior moved as predicted, but the chain hasn't reached churn yet | | Re-rank | Keep the Retention priority. Fix the weak link: stage the next home screen change to half of users so the churn link can be read as Causal | | Next Check | Churn at 120 days | | Failure Mode to Watch | Calling it Held because the 79% test result looked strong. That was a leading read; this is the outcome. | | Next Step Handoff | → glare-focus-decisions (Bets) to re-rank the remaining bets | ## 4. Prompt Strategies The prompts below show different ways to use this skill. Each example uses a mobile banking dashboard update. ### Prompt 1: Check Entry: Did the bet hold? *"Our mobile banking bet predicted sessions ending without an action would drop from 38% to under 30% in 60 days. It's day 60 and they're at 29%. Using the glare-lead-results skill, read up the chain, give the verdict, and tell us what to re-rank."* **Why this works:** It reads the outcome against the bet's own prediction, baseline, and window, not against how the test looked. **Best for:** - post-launch reviews - quarterly business reviews ### Prompt 2: Early Entry: Resist an early call *"Two weeks after launch, leadership wants to know if the new mobile banking home screen worked. Using glare-lead-results, tell us what we can say now, what we can't, and when the real verdict comes."* **Why this works:** It keeps launch-day excitement from being reported as the outcome. **Best for:** - pressure for early wins - launch updates ### Prompt 3: Miss Entry: When outcomes keep missing *"Our last three mobile banking bets were Partly held or Didn't hold. Using glare-lead-results, look for a pattern and tell us whether the problem is the designs or how the team works."* **Why this works:** Repeated misses usually point to how the team works, which the Design Assessment can rate. **Best for:** - teams with a string of misses - planning a process change --- *Glare Framework · glare-lead-results · Lead Area. Handoffs: glare-focus-decisions (Bets) · glare-lead-mapping · glare-lead-workflows · glare-measure-hunches (send a gap back) · glare-design-assessment*